What it does
Paste an email, a support ticket, meeting notes or a log file, and the redactor marks the personal details in it: names, email addresses, phone numbers, street addresses, postal codes, card numbers, IBANs, US Social Security numbers, Canadian SINs, dates of birth, IP addresses and API keys. You get a copy with each one swapped for a placeholder, ready to paste into ChatGPT, Claude or Gemini.
When the AI answers, paste its reply into step 2. The placeholders turn back into the real names and numbers, so you can send the result without retyping anything.
How it finds things
Most personal data has a shape. Card numbers pass a checksum (the Luhn check), IBANs have their own mod-97 check, phone numbers and postal codes follow national formats. The redactor uses those checks, so a random 16-digit order number isn't reported as a credit card. These matches are shown in the accent colour.
Names don't have a shape, so they're found from context: a greeting ("Hi Priya,"), a title ("Dr. Okafor"), a label ("Patient: …"), a sign-off, or phrases like "spoke to" and "1:1 with". Once a name turns up, every other mention of it is caught too, including the first or last name on its own. These are shown in amber so you can check them.
Anything it can't know about, like your company name, a project codename or a client mentioned only by first name, goes in Also hide these words.
Three ways to replace
| Style | Looks like | Use it when |
|---|---|---|
| Labels | [NAME_1], [EMAIL_2] | You want the AI to see clearly what was removed. Reversible. |
| Fake values | Alex Rivera, person1@example.com | You want the text to read naturally, for tone or grammar edits. Reversible. |
| Black bars | ████ | You're sharing a screenshot or a document and nothing needs to come back. |
The same value always gets the same placeholder, so the AI can still tell that the "Marcus" in paragraph one is the "Marcus" in paragraph four. Fake values come from reserved ranges (555-01xx phone numbers, example.com addresses, the 203.0.113.x test network) so they can never point at a real person.
Good times to use it
- Asking ChatGPT to draft a reply to a customer email
- Getting AI help with meeting notes, HR notes or an incident report
- Pasting server logs that contain user emails and IP addresses
- Sharing a support ticket in a public bug tracker or a Slack channel
Limitations
This is pattern matching, not mind reading. It won't catch a name that appears with no context at all, and it can't tell that "the only cardiologist in Moose Jaw" identifies someone. Read the safe copy before you paste it. Names are detected in English text; other formats are tuned for North America and the UK, with international phone numbers and IBANs also covered.
Privacy
Everything happens in your browser. The text you paste, the redacted copy and the list that maps placeholders back to real values stay in this tab. Nothing is uploaded or saved, and closing the tab clears it all.
FAQ
What shouldn't I paste into ChatGPT? Anything that identifies a person or lets someone into an account: names with contact details, government ID numbers, card numbers, health details, passwords and API keys. Our guide on what not to paste into ChatGPT goes through it with examples.
Can the AI still help if it can't see the names? Almost always. To write a polite refund email it needs to know there's a customer and an order, not who they are. The walkthrough on redacting personal information before using AI shows a full round trip.
Is an IP address personal data? Under GDPR it often is, and in a log file it can tie activity to a person. See what counts as PII for the grey areas.
Does it work with API keys too? Yes. It uses the same detection as our API Key & Secret Scanner, so a key in your notes gets redacted along with the personal details.